Analysis of Developers Prediction Performance on Software Models Using Back Propagation Algorithm

Abstract

That's what System Effort Estimation (SEE) is all about. Cost and schedule overruns are common in early phases of software development when the erroneous estimate is made. To build software development models, precise time and resource estimates are essential. Software models are typically built using a backpropagation neural network, that is the most prevalent architecture. Elman neural networks can be used in the same way as Back propagation networks. The difference between the predicted and actual effort in a successful prediction system should be as small as possible. Future training process will make use of historical NASA data. Back propagation has been found to be more effective in experiments than in Lstm neural networks. As the selections are made, the inputs' significance shifts. Comparing model test results with the model's predicted effort gives us an idea of the error rate in a model's predictions. With its help, a software project prediction model has been proposed.